Failure to control conventional cardiovascular risk factors in women with type 2 diabetes might explain worse mortality

The standardised mortality rate (SMR) for people with diabetes in England is 1.5–1.7, with differences in outcomes between sexes. There has been little work examining the factors that could have an impact on this or on what may determine sex differences in outcome.


| INTRODUCTION
[3] In England and Wales, a National Diabetes Audit in 2019 showed that the standardised mortality rate (SMR) for females was up to 26% higher than that for males, with the overall figure being reduced by the mix effect of a higher percentage of older females with lower SMR. 9We have recently shown that the SMR was higher in younger individuals with diabetes relative to those without diabetes.2][13][14] Nevertheless, the gender disparity in mortality outcomes persists. 15A meta-analysis of 43 studies, internationally, has found lower prevalence of aspirin, statins, and angiotensin converting enzyme inhibitors prescription to women at high cardiovascular risk in primary care. 16Likewise, studies in secondary care have demonstrated that women are generally less likely than men to have a prescription of evidence-based cardiovascular medications after a cardiac event. 17,180][21] We have recently suggested that one driver for sex-difference in cardiovascular disease outcomes in females with type 2 diabetes is due to inequitable use of newer therapies with proven cardiovascular benefit. 10Differences may also result from failure to attain targets for conventional cardiovascular risk factors and there are few data for how cardiovascular risk factors and mortality risk change over the time course of type 2 diabetes, between sexes.
We have therefore determined the presence of cardiovascular risk factors, and prescribing of therapies with cardiovascular benefit, in males and females with type 2 diabetes in the United Kingdom (UK) population.We have modelled the effect of these factors on SMR.

| Patients
Fully anonymised data for individuals with a diagnosis of T2D, recorded in the Salford Integrated Record (SIR; Salford, England), 22 was extracted for a 2010 cohort which was inclusive of individuals newly diagnosed with T2D thereafter.Data were followed, until the end of January 2020, or death.The data included the date of birth, date of T2D diagnosis, body mass index (BMI), test results, the medication prescribed and date of death.
The data extract was approved by SIR Governance Board (ref: DATA022).

| Measures of impact of sex on mortality
Three outcomes were used to evaluate the impact of sex on mortality.

| Standardised mortality rate (SMR)
The annual expected deaths by age and sex for the general population were taken from the annual Office of National Statistics (ONS) mortality rate.The ratio of the sum of expected patient deaths with the actual deaths provided the SMR.

| Standardised mortality rate adjusted for deprivation (SMRd)
There is a strong established link between local social deprivation and mortality rate and Salford has higher social deprivation than the national average, so an adjustment was made to create a national comparison.The Office of National statistics provides, by age and sex, the recorded deaths and population, and the local Index of Multiple Deprivation (IMD) for the 37,400 Lower Layer Super Output Areas (LSOAs) in England.Simple linear regression was used to derive the factor relating IMD and SMR to each LSOA.As all patients reside in a given LSOA, this factor was applied to adjust the mortality for that LSOA back to the national median IMD.

| Life expectancy years lost/death (LEYLD)
The ONS life expectancy value for each actual death, given their age and sex, was summed and from this was subtracted the total of fraction of expected deaths multiplied by the life expectancy values, to provide a total net life expectancy year lost.This number was then divided by the number of actual deaths to give the LEYLD.This is  To assess the potential impact of prescriptions for diabetes and cardiovascular risk reduction (such as statins and anti-hypertensives), the history of prescribing was identified.We used a minimum limit of 12 prescriptions, reflecting one year of use.As specific measures were often not recorded in the actual year of death, the last recorded value was used.

| Statistical analysis
Data were assembled within Excel (Microsoft corporation, 2018) for each patient year.The SMRd, LEYLD, OR and hazard ratio (HR) were calculated directly for each cohort and then linear and multifactorial logistic regression was applied to the individuals from selected cohorts using the statistical Analyse-It add in (Leeds, UK).
To establish the size of the effect, analysis reflected several questions: 1) How much do risk factors vary between sexes?The prevalence values of risk factors were established for the two groups split by sex, and the odds ratio calculated female versus male.
2) How much variation was there between sexes in the patient numbers by age, age at diagnosis and duration with T2D?The percentage of the total number of patient years recorded for each age and duration with T2D and the percentage of total patients by age at diagnosis were compared between the sexes.

3) How does mortality risk develop over duration with disease?
The development of cumulative hazard ratio taking into account the duration of T2D for each sex as a ratio to the expected cumulative mortality rate for the general population of the same age and sex.

4) Do the prevalence and clinical outcomes vary across age and
duration, to a different degree, between the sexes?Heat maps were prepared across age at diagnosis year groups and duration with diabetes year groups to highlight the differences in prevalence and outcomes for females and males.

5) How do cardiovascular risk factors develop within sexes over the course of living with T2D?
The absolute level of risk factors was compared between sexes over the duration with T2D.

6) Does the impact of risk factors on mortality differ by sex?
Multifactorial logistic regression of risk factors present in the last recorded year of each patient was linked to actual deaths for both males and females to calculate the OR level for each of the risk factors by sex.

7) How does the prevalence of risk factors in each sex compare with the level of mortality risk associated with each factor?
Univariate ORs were calculated from the prevalence of risk factors in each sex and from the prevalence of risk factors in those who died versus.those who survived.

8) What is the overall impact of modelling the (female) prevalence of cardiovascular risk factors to the equivalent levels seen in the male population on total female deaths?
The potential reduction in female death was calculated by applying the male risk factor prevalence to the female numbers and applying the common mortality risk associated for those with and without the factor.

| By how much do the outcomes and risk factors vary between the sexes?
Table 1 provides an overview of the study cohort.Data from n = 11,806 (female 5184; male 6622) were included.Of these, n = 5540 were newly diagnosed (F = 2375; M = 3165) and n = 3921 ).Thus, the SMRd and LEYLD were higher in females (F/M ratio 1.12 and 1.18, respectively).

| How much variation was there between the sexes in the patient numbers by age, age at diagnosis, and duration with T2D?
The overall age distribution was similar between sexes (Figure 1A) although there was a male preponderance in the 40-70 age group, while females were more heavily weighted in >70 age group.Fewer females were diagnosed between ages 40-60 than males (Figure 1B).
Whereas over the age of 70 years, a higher proportion of females was diagnosed than males.With the earlier death of males (Tables 1 and   2), the net result was that males and females had diabetes for a similar length of time (Figure 1C).

T A B L E 1
Population split between male (M) and female (F).

| Does the mortality risk develop differently by sex with duration with disease?
Females had a consistently greater cumulative hazard ratio for mortality than men.This was apparent from year 2 (ratio 1.17, compared to 1.05 in males; Figure 2) and in females, rose to a peak ratio 1.37 after 10 years at an average age of 70.2, while males rose to a peak of 1.30 after 14 years with an average age 70.1 years.Both then slowly declined over the longer term as the expected general mortality rate increased.

| Do the prevalence and clinical outcomes vary across age and duration to a different degree, between the sexes?
To examine the interaction of age at diagnosis and duration of living with T2D, we calculated the SMRd for each age group and duration of diabetes (Tables 3 and 4).The higher OR for death in females was most pronounced in those of a younger age at diagnosis but with a longer duration of diabetes (e.g., OR 2.12 for those diagnosed between 40 and 49 years of age and who had duration of diabetes 15-19 years).Greater female risk was mitigated in those with diabetes over 20 years duration and in advanced age, possibly due to a survivor effect.
We estimated that of the 2080 recorded deaths for males with T2D, n = 525 (25%) were excess to those that would be normally expected, and in females n = 598 (32%) of the 1841 recorded females were excess (Table 4).
In Tables 1 and 2, if one simply compares the overall SMRd for females (1.60) with those of males (1.44), one could assume that 25% (190) of the 689 excess T2D deaths in females might be linked to their sex.However, Table 3 shows that the proportion of females diagnosed at an older age is higher than that of men, and the difference in SMRd versus non-diabetic women is much lower.If the SMRd differences are reconciled by age groups as shown in Table 4, the total excess females when the male SMRd is applied is 310 or 45% of the 689 total estimated excess deaths in females with diabetes.

| How do cardiovascular risk factors develop within sexes, over the course of living with T2D?
Figure 3 shows how the level of risk factors with the most variation changes over the duration of T2D.A greater proportion of females who were living with diabetes for less than 15 years had been diagnosed over the age of 65 years (Figure 3, Panel A).This sex difference was lost in those surviving with diabetes for >15 years.Figure 3B shows that the proportions with HbA1c >58 mmol/mol (7.5% DCCT units) were similar between sexes.In each case, the proportion with HbA1c >58 mmol/mol increased with increasing duration of diabetes.
Figure 3C shows that a greater proportion of females had eGFR <60 mLs/min/1.73m 2 than men.The proportion of women, at any age, with eGFR <60 mLs/min/1.73m 2 was similar to that seen in men but who were 5-10 years older (up to 30 years after diagnosis).This value is greater than the 2.4 year older (mean) age in women (Tables 1 and   2).A greater proportion of males are prescribed newer glucose

| Does the impact of risk factors on mortality, differ by sex?
Multivariable logistic regression showed that the OR for the risk factors was similar between males and females (Figure 4).
The only risk factor with a significant sex difference was 'history of sulphonylurea prescribing'-where prescribing of sulphonylurea showed a greater mortality disadvantage to women (OR 2.5, compared to men with OR 1.6).However, overall, a lower proportion of women with diabetes (0.94) were prescribed sulphonylureas (Figure 5).

| Does the prevalence of risk factors in each sex change with respect to factors that increase or decrease mortality risk?
The univariable ORs for the prevalence of cardiovascular risk factors in females versus males (in blue) as well as the ORs for cardiovascular risk factors in those who died versus survivors (in red) are shown in and 'history of SGLT2-i prescribing'.

| Modelling the impact of applying male prevalence of risk factors to the female population, on total female deaths
Table 5 shows the predicted effect on mortality if the prevalence of any one of the risk factors in females was the same as that in males.
The total estimated excess female deaths, accounting for the factors in Table 5, were n = 437.Although this was a univariable analysis, theoretically up to 437 lives saved could be achieved, although given the likely interdependence of risk factors, the true total may be lower than that estimated.This figure is similar to the 310 additional deaths identified earlier through the SMRd difference.

| DISCUSSION
We have described how a differential mortality rate continues to impact people with T2D in relation to the general population.We show that the prevalence and clinical outcomes vary across age and duration to a different degree between the sexes.Specifically, our analysis highlights that females with T2D are at particularly greater mortality risk and that this is more pronounced at younger age.The hazard ratio of death (actual mortality/expected mortality rate) was Cumulative hazard ratio for death after a diagnosis of type 2 diabetes.
higher in females, even in the early years after diagnosis and reached a maximum difference at around 10 years after diagnosis.A differential effect for cardiovascular mortality with respect to younger age has been shown in a meta-analysis.In that study, women with diabetes aged 35-59 years had a relative risk of 5•55 for occlusive vascular mortality. 8Furthermore, we found that women were diagnosed with diabetes, on average, 2-years later than males.Prolonged duration of pre-diabetes, 23 and later age of diagnosis of T2D in females have been noted previously. 24This is unlikely to be due to deficient screening for diabetes in women -as women appear to have greater participation in diabetes screening programmes. 25Sex differences in body anthropometry likely contribute to the later age at diagnosis in women.There is a menopausal-related increase in total body fat, and distribution to the intra-abdominal area, 26 with the result being that, by the seventh-decade of life, women have equivalent prevalence of T2D to men. 27 has been suggested that, given the broadly favourable metabolic milieu of pre-menopausal women, for a woman to develop diabetes she must have accrued a more extensive metabolic insult than men. 28This is consistent with data showing a 2-year T A B L E 3 Heat map across the age of diagnosis and duration with T2 for population with T2 and then split between male/female for standardised mortality rate (SMR) and the odds ratio.greater duration of pre-diabetes in women 23 and that, when first diagnosed with diabetes, women have a BMI nearly 2 kg/m 2 higher than that of men, despite similar levels of HbA1c. 29,30Biological explanations can also account for the difference in CVD mortality as the menopause leads to the loss of the protective effect of oestrogen or the balance of oestrogen and androgens. 31Diabetes eliminates the hormonal protective CVD benefit in pre-menopausal females and is also likely to further exacerbate CVD risk in postmenopausal female further, compared to female without diabetes. 31Alternative explanations include that females with diabetes are less likely to be physically active, have lower use of aspirin, ACE-inhibitor (as seen here) and beta-blockers, 32 and may require greater frequency and intensity of physical activity to reduce their cardiovascular risk. 33derstanding the impact of the factors that render females with T2D at greater relative mortality risk than males, can help to facilitate the changes in management necessary to reduce this disparity.

Years with T2D
Our multivariable logistic regression showed that the odds ratio for the effect of individual risk factors was similar between males and females, suggesting that the underlying metabolic processes are not different between sexes.This is in keeping with a meta-analysis showing that total cholesterol, blood pressure, and BMI each had continuous log-linear associations with occlusive vascular mortality that were similar in strength among those with and those without diabetes irrespective of sex. 8Greater efforts must therefore be made to control conventional cardiovascular risk factors.For instance, a recent evaluation of data from Pacific countries showed that men were 50%-60% more likely to attain at least three treatment targets (HbA1c <7%, blood pressure <130/80 mmHg, risk-based LDLcholesterol, and lack of central obesity).Our data suggest that the optimisation of known risk factors would come close to ending the gender gap in cardiovascular outcomes.However, an alternative hypothesis is that women may be more at risk from novel cardiovascular risk factors, such as inflammatory markers, and this needs further research. 21ile the nature of the risk factors for mortality was not different in females than males, our analysis found that the higher proportion of females with a low eGFR (<60 mL/min/1.73m 2 ) may contribute to the higher mortality rate in females.The difference between the proportion of women and men with eGFR<60 mL/min/1.73m 2 at any age (with men showing a similar proportion 5-10 years older) (Figure 3C) may relate to the prolonged period of undiagnosed diabetes 23 or to inherent sexual dimorphism of the metabolic and molecular mechanisms underlying diabetic kidney disease. 34We suggest that this difference in renal deterioration outweighs the difference in chronological age of 2.4 years.Women are at greater risk of developing endstage renal failure in diabetes -having a 38% greater relative risk of ERSF compared to men. 34,35Fewer prescriptions of reno-protective SGLT2-i and ACE-I or ARB may contribute to this, 36,37 but so too a sex difference in response to ACEI and ARB.

| Strengths/limitations
This analysis only covers one region of the UK.However, Salford is considered to have had a well-run diabetes service in both primary and secondary care.The recorded diagnoses, medication, blood test results and anthropometric measures encompass all that is held in general practice records for the individuals in this study.Individuals who have left the area will not have follow-up data available for analysis; therefore, there is the potential for some health events/ deaths to be missed.We also accept that certain factors may not be independent; for example, older age, eGFR, UACR, ACE-I/ARB prescription, and SGLT2-i prescription may be interrelated.We did not have data for the cause of death.A meta-analysis of over 5 million people has suggested a sex difference in coronary heart disease as well as all-cause mortality. 15Finally, we did not have data on the prescription of hormone replacement treatment (HRT) in menopausal women with T2D.

| CONCLUSION
We suggest that greater clinician awareness of the higher mortality rate in females with T2D, coupled with the greater provision of interventions to improve diabetes processes of care, is needed to decrease the sex gap in mortality rates.In particular, these interventions are needed in younger women and considerable effort must be made to diagnose women with diabetes, at an earlier age.
T A B L E 5 Number of excess female deaths estimated when the proportion of any one cardiovascular risk factor attainment was equivalent to that of males.

Factor
described by the equation: (Σ (Actual deaths � Life Expectancy (for age & sex)) −Σ (Expected deaths � Life expectancy (for age & sex)))/ Actual deaths.Two aspects of the impact of patient's age were considered separately: the age at diagnosis and the duration with T2D, in years.

F I G U R E 1
Distribution of patients by sex.Panel 1A, patient age; Panel 1B, age at diagnosis; Panel 1C, duration of type 2 diabetes.lowering drugs and ACE-i/ARB, irrespective of the duration of diabetes (Figure 3; Panels F-H).This was also the case with statin therapy (Figure 3; Panel E), despite a greater proportion of women not achieving total cholesterol <5 mmol/L (Figure 3; Panel D).

F I G U R E 3 F I G U R E 4
Cardiovascular risk factors develop over duration with type 2 diabetes.(A) % age at diagnosis >65.(B) % HbA1c >58 mmol/mol.(C) eGFR <60 mL/min/1.73m 2 .(D) TC <5 mmol/L.(E) % on statins.(F) % on ACE-I/ARBs.(G) % history of DPP4-i.(H) % history of SGLT2.Multifactorial logistic regression for males and females linking reported deaths to odds ratio for major risk factors at last recorded year.F I G U R E 5 Univariable odds ratios for females (3,344) compared to males (4585; blue colour) (at last clinical record) as well as for deaths (3,847) compared to survivors (7958; red colour).
Heat map of excess type 2 diabetes (T2D) deaths by age at diagnosis and duration with T2D for females and males plus the number of those excess deaths for females above the standardised mortality rate deprivation (SMRd) = males.
T A B L E 4STEDMAN ET AL.